Help in neural network (newff) function

[Digital_11]

Hi everybody,
If I load mat files and pass them to newff function as parameter, the newff function does not work,

net= newff(trainMFCC, TrainTargets, 20);%trainMFCC=load(trainMFCC.mat)
net = train(net,trainMFCC, TrainTargets);%TrainMFCC=load(Target_Train.mat)
Outputs = sim(net, testMFCC);%testMFCC =load(Target_Test.mat)
where the mat files contains the extracted voice feature,
and the error is
??? Error using ==> newff>new_5p1 at 154
Inputs are not a numerical or logical matrix or cell array.

Error in ==> newff>create_network at 130
net = new_5p1(varargin{:});

Error in ==> newff at 102
out1 = create_network(varargin{:});

can any body help please,

 

[lebesgue]

Are you working on speaker recognisiton using neural networks? I ask this because in your code MFCC was cited.. I am also working in the same problem..

 

[bigdogguru]

Hi,

While I'm not familiar with the newff() or the Neutral Network Toolbox, I am quite versed in advanced Matlab programming and would be more than happy to take a look at your project.

Looking at the small code snippet and error messages provided, it would appear to be a formating issue, but without seeing the code in its entirety and data in the .mat files there's no way to be sure.

If you would upload a zipped copy of your project, I'll attempt to debug it for you.

Have you successfully used newff() in the past? If so could you provide a short example of its use and desired output as well?

Ciao,
BigDog

 

[Digital_11]

Thanks to everybody, the problem partially solved ,you can find usefull link bellow,
**broken link removed**
**broken link removed**

Are you working on speaker recognisiton using neural networks? I ask this because in your code MFCC was cited.. I am also working in the same problem

Yes.

 

[lebesgue]

Hi all . I am also working in the same problem as Digital_11.
I have generated three matrices w1,w2,w3 for three speakers which are the MFCC coefficients. The order of the matrix is (100x10). 100 is the number of frames and 10 is the number of cepstral coefficients. How to form the training input data for the network? I am using nntool for implementation. I don't know to form the training input vectors. please help me. In digital_11 program I am very much concerned about trainmfcc.mat